Interrelationship of alcohol misuse, HIV sexual risk and HIV screening uptake among emergency department patients

From July 2009 to August 2009, 18- to 64-year-old ED patients were randomly selected for inclusion in this study. This investigation had two components: (1) a cross-sectional study examining the prevalence of alcohol misuse and HIV sexual risk among ED patients, (2) and an examination of opt-in HIV screening in this population. The study was conducted at two academic EDs (Rhode Island Hospital and The Miriam Hospital) located in Providence, Rhode Island, that are affiliated with the Alpert Medical School of Brown University. Rhode Island Hospital is a level 1 trauma center, receiving over 100,000 annual adult patient visits, and the Miriam Hospital is a community hospital, receiving over 55,000 annual adult patient visits. The Rhode Island Hospital Institutional Review Board approved this study. Verbal consent was obtained for the cross-sectional component of the study, and written consent was obtained for the HIV testing component.

Similar to previous studies conducted in these EDs [44, 70], this study obtained a representative sample of participants by approaching randomly selected ED patients for study inclusion on randomly selected dates and shifts. Fifty-seven dates during an eight-week period (July 2009-August 2009) were randomly selected; all days of the week had an equal chance of selection. For these 57 dates, 72 shifts were randomly selected using a weighting scheme corresponding to patient ED volume during a typical 24-hour period (40% of shifts were from 8:00 am-4:00 pm, 50% from 4:00 pm-midnight and 10% from midnight-8:00 am). On those shifts, 80% of patients in the ED were randomly selected for possible inclusion in the study. This random selection of patients was based upon their ED medical record number, the last two digits of which were matched to numbers randomly selected by a computer program (http://​www.​random.​org). Patient eligibility for the study was assessed for these randomly selected patients by a research assistant (RA) through a review of their ED medical record and confirmation of their eligibility through an in-person interview. Patients were study eligible if they were: age 18-64-years; English- or Spanish-speaking; not critically ill or injured; not prison inmates, not under arrest, or undergoing home confinement; not presenting for an acute psychiatric illness; not intoxicated; not HIV infected; not participating in an HIV vaccine trial, and did not have a physical disability or mental impairment that prevented them from providing consent for participating in the study. No incentives were offered to participants. ED staff members were not permitted to encourage or refer patients to be in the study.

Participants were interviewed by the RA about their demographic characteristics (age; race/ethnicity; partner status; insurance status; and education level) and history of ever being tested for HIV through blood donation, screening, or diagnostic testing; and time elapsed since blood donation or HIV testing. These demographic and HIV testing history questions were developed for and used in previous studies [34, 62, 70]. Participants completed self-administered confidential questionnaires regarding the quantity and frequency of their alcohol use, severity of their alcohol use, and sexual risk for HIV on tablet computers using the Questionnaire Development System (QDS) (NOVA Research Company, Bethesda, MD). The survey questionnaires were finalized in English, translated into Spanish then back translated into English to ensure translation accuracy using accepted techniques [71‐74]. An English-language copy of the questionnaires is provided in the supplementary material (see Additional file 1). The questionnaires were available in English or Spanish and were completed by participants while they awaited medical care.

Alcohol misuse and disorder severity during the past 12 months was measured through the ten-item Alcohol Use Disorders Identification Test (AUDIT), developed by the World Health Organization [23]. The AUDIT is a well-established alcohol misuse screening and severity instrument for the ED and other settings with excellent reliability and validity [23, 75‐77]. Quantity and frequency of alcohol use during a typical month in the past 12 months was assessed by a six-question survey (The Alcohol Use Questionnaire) developed by the study authors for the purpose of this study and based on research questions used in previous studies [23, 78, 79]. Through this questionnaire, participants were queried about the number of days they spent drinking in a typical month, the number of drinks consumed on a typical day, their alcohol beverage choice, the most number of drinks consumed on one occasion, and the number of days spent engaging in binge drinking in a typical month. Binge drinking was assessed using NIAAA recommended definitions [16]. Per these recommendations, male participants were asked on how many days they consumed five or more drinks and female participants were asked on how many days they consumed four or more drinks on one occasion in a typical month during the past 12 months. The Alcohol Use Questionnaire complemented the AUDIT in that participants were asked for specifics regarding the number of days they drank alcohol and the amounts used; whereas, the AUDIT employs categorical designations as the responses for these questions. Further, the Alcohol Use Questionnaire permitted sex-specific responses for binge drinking. A Chronbach’s Alpha analysis showed an acceptable level of internal consistency (α = 0.80), and a strong correlation between relevant questions from the Alcohol Use Questionnaire and total AUDIT scores (ρ = 0.66-0.73) in the population included in this study.

Participants also completed the HIV Sexual Risk Questionnaire, consisting of multiple-choice, closed-response questions about their reported HIV sexual risk behaviors. The questions were derived from the CDC National HIV Behavioral Surveillance (NHBS) System survey and adapted through cognitive testing for this study and previous studies [34, 43, 70, 80]. This questionnaire consisted of primary questions with associated sequences of follow-up questions, which would only appear if the participant answered affirmatively to the primary questions. The number of questions answered by each participant was dependent upon their reported HIV sexual risk behaviors in the past 12 months. Sexual risk for HIV was assessed separately for males and females due to the different types of sexual risks they engage in; therefore, the questions were sex-specific. Accordingly, females were asked questions regarding anal and/or vaginal sex with males and males were asked about anal and/or vaginal sex with females, and anal sex with males. A Chronbach’s Alpha analysis confirmed a strong level of internal consistency for these questions among females (α = 0.90) and among males (α = 0.84).

In the HIV Sexual Risk Questionnaire, participants were queried about HIV sexual risk behaviors in the past 12 months by sexual partner type as by CDC-recommended definitions [80]. Sexual partner types were: (1) main partner(s), defined for participants as “men or women you felt committed to, such as boyfriends or girlfriends, husbands or wives, significant others or life partners”; (2) casual partner(s), defined as “men or women you had sex with but did not feel committed to”; and (3) exchange partner(s), defined as “men or women you gave money, drugs, or other things to pay for sex, or men or women you had sex with so they would give you money, drugs or other things.” Participants were asked to report the type(s) of sexual partners (main, casual and exchange partner), unprotected sex (anal and/or vaginal sexual intercourse) with these partners and the number of sexual partners by partner type according to their partner’s history of injection drug use, sexually transmitted diseases (STDs), and HIV status. According to each partner type, participants were asked how many of their sexual partners they (1) knew or (2) how many they were unsure if they had HIV, (3) were injection-drug users, or (4) had an STD. In addition, females also were asked the number of male sexual partners they knew and also the number of male sexual partners they were unsure about had had sex with other males.

The study authors developed three additional questions regarding the intersection of alcohol misuse and HIV sexual risk behaviors. The questions asked participants if they had ever had sex while intoxicated, regretted ever having had sex while intoxicated, and if they were ever unsure if they had sex while intoxicated in the past 12 months. A Chronbach’s Alpha analysis confirmed an acceptable level of internal consistency for these questions among female drinkers (α = 0.73) and among male drinkers (α = 0.67).

At the conclusion of the study, participants were asked by the RA if they would like to be tested for HIV using a free rapid HIV test (opt-in HIV screening). Participants were informed that HIV screening was voluntary, involved a rapid HIV test using a finger stick of blood, and that results would be provided to them within 20–30 minutes. The OraQuick ADVANCE® Rapid HIV-1/2 Antibody Test was performed (http://​www.​orasure.​com). No incentives for HIV screening were offered and participants were not provided with an intervention or encouragement to be tested. RAs were blinded to participant’s alcohol use or misuse and HIV sexual risk history. Participants were not informed at the start of the study that they would be offered an HIV test. Uptake of HIV screening was an outcome measure for the study. As such, the relationship of participant reported alcohol misuse and sexual risk for HIV to uptake of HIV screening was assessed. A follow-up question asked participants about the main reasons why they accepted or declined screening. All participants who agreed to be tested were offered HIV risk-reduction counseling in English or Spanish. All RAs were bilingual in English and Spanish. The RAs also completed a state-sponsored training program on HIV counseling and screening as well as a training program on rapid HIV screening. No patients tested positive for HIV in this study.

All statistical analyses were conducted using STATA 11 (Stata Corp., College Station, TX). Participant screening and enrollment were summarized and diagramed per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations [81]. Demographic characteristics, HIV screening history, alcohol misuse, sexual risk for HIV and reasons why participants accepted or declined HIV testing were also summarized by sex. Data are reported using mean, median, standard deviation (SD), and interquartile range (IQR) where appropriate.

The percentage of days spent drinking in one month was calculated by dividing the number of days spent drinking in one month by 31 days. Presence of binge drinking was determined by the aforementioned cutoffs of ≥ five drinks for men and ≥ four drinks for women [16]. The percentage of days spent binging was calculated by dividing the number of days spent binging by the number of total days spent drinking in one month. Percentage of days spent drinking in one month and percentage of days spent binging were converted into four levels (0-24%, 25-49%, 50-74%, 75-100%) because values were not normally distributed and to aid in ease of interpretation. For the AUDIT, participants were classified into at-risk drinking levels as recommended by Babor, Biddle, Saunders and Monteiro [23]. For men, a score of < eight for men and < six for women indicated a low-risk drinking level [82]. A score ≥ eight and ≤ 15 for men and ≥ six and ≤ 13 for women indicated a hazardous drinking level [83]. A score ≥ 16 and ≤ 19 for men and ≥ 14 and ≤ 17 for women indicated a harmful drinking level. A score of ≥ 20 for men and ≥ 18 for women determined a dependent drinking level.

Based upon their HIV Sexual Risk Questionnaire responses, all participants who reported no sexual intercourse within the past 12 months were eliminated from the study analysis. HIV sexual risk scores were calculated for those who reported having sexual intercourse within the past 12 months. Points for HIV sexual risk scores were assigned based on the reported type of sexual partner. We assigned one point for participants who reported having unprotected sex with their main partner, and two points each for having unprotected sex with a casual partner and/or with an exchange partner. Additional points were assigned based upon the number of unprotected sexual partners and upon characteristics of the participant’s sexual partners (e.g. HIV status, injection- drug use and history of STD infection). The highest possible score was 209 for females and 514 for males. HIV sexual risk scores were transformed into a logarithmic scale because the scores were not normally distributed. The log of HIV sexual risk scores were divided into tertiles for ease of interpretation. Tertiles were based upon the distribution of the log of the HIV sexual risk scores, and not an even distribution of the participants. As such, they were grouped according to three levels of reported HIV risk, and hence the sizes of the groups were not equal.

Multivariable and univariable regression analyses were used to assess for relationships between (1) log of HIV sexual risk scores in tertiles and alcohol misuse, (2) HIV screening uptake and alcohol misuse, (3) HIV screening uptake and sexual risk for HIV, and (4) HIV screening uptake and the intersection of HIV sexual risk and alcohol misuse (sex while intoxicated, regret ever having had sex while intoxicated, and unsure if ever had sex while intoxicated). Ordinal logistic regression modeling was performed for analyzing associations between the log of HIV sexual risk levels in tertiles and whether participants drink or not; percentage of days spent drinking and binging in one month; AUDIT at-risk drinking levels; and whether participants binge or not. Logistic regression modeling was used to assess the outcome of HIV screening uptake as related to (1) alcohol misuse, (2) the log of HIV sexual risk levels in tertiles; and (3) the intersection of alcohol misuse and sexual risk for HIV. Based upon responses for declining HIV screening, logistic regression modeling assessed the outcome of participant’s perception of not being at risk for an HIV infection and drinking and binging status among all participants and drinkers. Goodness-of-fit of the logistic regression models was confirmed by Hosmer-Lemeshow analyses. Adjusted odds ratios (AORs) with corresponding 95% confidence intervals (CIs) were estimated. Multivariable regression models were adjusted for participant demographic characteristics (age, race/ethnicity, partner status, insurance status and education level). Our previous research indicated that demographic characteristics are important correlates for uptake of HIV screening, hence we adjusted for our main effects for these confounding variables [43, 62]. All analyses were considered significant at an α level of 0.05, with no adjustments for multiple comparisons.

During the two-month study period, 2,565 randomly selected 18-64-year-old English or Spanish-speaking ED patients were assessed for study eligibility. Of the 887 study eligible ED patients, 750 enrolled in the study. Figure 1 depicts the results of eligibility assessments, the major reasons for study ineligibility and for accepting and declining study participation. As shown, 28.9% of participants reported not having sexual intercourse in the past 12 months, which left 524 participants who reported some sexual risk for HIV, and who constituted the final study sample used for these analyses.

Table 1 shows the participant demographic characteristics and HIV screening history by sex. Of all 524 participants, 55.0% were female. The median age was 39 (IQR, 27–50 years) for males and 34 (IQR, 26.5-45.5) for females. For both males and females, most participants were white, non-Hispanic, had private health insurance, had never been married, and had 12 years or more of formal education. Seventy-two percent of participants reported having been tested for HIV (76.4% of women, 66.5% of men). The majority of participants in this study had been tested for HIV more than five years ago, had never donated blood, and had been tested for HIV but not as part of a blood donation.

Table 2 depicts participant alcohol misuse and sexual risk for HIV by sex. Approximately 75% of participants reported drinking alcohol within the past 30 days.

Thirty-three percent of males and 16.3% of females reported spending 75 – 100% days of the month drinking alcohol. The majority of participants fell within the low-risk drinking level based on AUDIT scores; however, 74.5% of males and 59.6% of females reported binging, and 26.1% of both females and males would be classified as drinking at hazardous levels. The majority of participant’s log HIV sexual risk scores fell within the first tertile level. The majority of participants, 86.1% of females and 68.2% of males, reported having only a main partner. Among drinkers, 88.6% of females and 79.8% of males reported having unprotected sex, and among non-drinkers, 80.0% of females and 80.9% males reported having unprotected sex in the past 12 months. Approximately 36.2% of males and 25.1% of females reported having sex while intoxicated, 15.4% of males and 12.3% of females regretted ever having had sex while intoxicated and 6.4% of males and 5.0% of females were unsure if they had sex while intoxicated in the past 12 months.

In Table 3, results of multivariable logistic regression analyses demonstrate, for both males and females, drinking status (whether participants drink or not), was strongly associated with an increase in sexual risk for HIV. For female drinkers, a greater percentage of days spent drinking and binging, reaching hazardous and harmful AUDIT levels and binging status were associated with increased sexual risk for HIV among females. For male drinkers, a higher percentage of days spent drinking and hazardous and dependent AUDIT levels were associated with increasing sexual risk for HIV.

As shown in Table 3, multivariable logistic regression was performed to assess the relationship between HIV screening uptake and alcohol misuse. There was no relationship found for both males and females. Table 4 illustrates the results of the multivariable logistic regression analyses examining the association between HIV testing uptake and log of HIV sexual risk, stratified by alcohol misuse. No relationships were identified among all participants, participants who drink alcohol and do not drink alcohol, for both males and females.

Univariable logistic regression and multivariable logistic regression analyses were conducted to assess the associations between HIV screening uptake and the intersection of sexual risk for HIV and alcohol misuse, as shown in Table 5. Univariable regression analyses revealed a strong relationship between HIV screening uptake, regretting ever having had sex while intoxicated, and unsure if ever had sex while intoxicated among female drinkers. For male drinkers, a relationship was found between HIV screening uptake, sex while intoxicated and unsure if ever had sex while intoxicated. However, when adjusting for demographic characteristics in the multivariable logistic regression analyses, there were no relationships found for both males and females.

We examined factors related to reasons why participants accepted or declined HIV screening in the ED. Of those who agreed to screening, among women, 33.7% of drinkers and 25.5% of non-drinkers cited convenience as the most common reason why they accepted screening. Among men, 26.9% of drinkers and 32.1% of non-drinkers who accepted screening cited “because you asked” as the most common reason. Of the participants who declined screening, among women 51.7% of drinkers and 60% of non-drinkers, and among men 46.9% of drinkers and 63% of non-drinkers cited they did not believe they were at risk as the most common reason for not being screened. In examining the relationship between alcohol misuse and acceptance of screening, alcohol drinkers were just as likely as non-drinkers to say that they were not at risk for HIV among males (AOR 2.33 [0.89-6.11]) and females (AOR 0.83 [0.35-1.94]). Bingers were just as likely as non-bingers to say that they were not at risk for HIV among males (AOR 1.50 [0.62-3.64]) and females (AOR 1.03 [0.41-2.63]).

This study had a number of limitations. Self-report data regarding alcohol consumption and sexual risk for HIV may be inaccurate. Study participants may have underestimated or not recalled information regarding their alcohol consumption and HIV testing history. However, self-report of alcohol consumption and sexual behavior can be a reasonable method of obtaining these data [94, 95]. Also, we did not collect data on whether or not the participant’s ED visit was related to their alcohol use. We do provide information regarding their level of at-risk drinking. Social desirability factors may have influenced some patients in their responses to reasons for accepting or declining screening, rather than any perception of their risk. Furthermore, it is unclear whether acceptance of screening based on an opt-out approach in the ED would be similar for participants who were excluded from the study. However, an opt-out approach may not be appropriate for patients who are unable to provide study consent. The HIV Sexual Risk Questionnaire has not been validated as a predictor of acquisition of HIV. As such, the true relationship between reported risk and HIV acquisition cannot be determined by this study. In addition, only 15.2% of women and 29.3% of men reported having unprotected sex with a casual partner (with or without an exchange or main partner), and most participants reported only sex with a main sexual partner. As such, the majority of participants could potentially be considered at lower risk for acquiring HIV, which might have appropriately influenced the uptake of testing. The small sample size may have produced limitations in identifying differences when they do exist. The study outcomes may not be appropriate for other EDs with different demographic characteristics, even though we attempted to obtain a representative sample by randomly selecting dates, shifts and participants. Of course, a failure to demonstrate relationships, should they exist, could have been related to the nature of the questions asked, their format, and the interpretation of the questions by the participants. Alternative questions, topics, and approaches could yield different results.